Removing Moiré Patterns

Moiré patterns are one of most annoying things that happen when working with scanned bitmap images. You see them every now and then especially when you scan a printed originals (as opposed to continuous-tone images like photos).

The physical nature of Moiré lies in interference of two or more regular structures with different special frequencies. You can see this effect in real life if you go past two fences located one behind other or when you look at folded silk stockings.

In order to get the idea of Moiré origin, take a look at the following illustration:

Moire pattern 1

The two linear patterns have different frequency and being superimposed they will show frequency beats as could be easily seen on the illustration. Similar effect appears when overlaying two dot grids with regular structure being rotated with respect to each other:

Moire2.GIF (3698 bytes)

When you scan an image it is digitized by a line of photosensitive elements. The number of elements that sense light coming from the image is determined by the optical resolution of the scanner and the size of an image. For example, if you scan with the 600 dpi scanner and the image is 3" wide, then 1800 photoelements will take part in digitizing the image. Each photoelement is a discrete unit. Here I should note that specifying the scan resolution other than optical resolution of the scanner requires additional mathematical operations to upsample/downsample the resulting image scanned. If resampling is performed by an integer fraction (e.g. 2:1 or 1:4) the mathematical transformations cause less distortion than if you resample it by a fractional scale (2:3, 1:1.5 and so on). Therefore it is a usual rule for scanning to scan at integer scale to reduce quality loss.

When the image has a regular pattern (e.g. a jacket with fine fabric weave), it is superimposed with the scanner grid (the line of photoelements) that may cause interference and result in moiré patterns. This is especially true if you scan printed originals because usually a continuous-tone image is printed with halftone screens. The screen frequency and a scanner resolution may mismatch producing those dreaded moiré patterns. A misaligned or rotated original produce them too. If your scanned image reveals heavy moiré effect, try rotating the original in the scanner. Experiment with the angle of rotation to achieve the minimal moiré pattern. Scan at the highest possible resolution. Even if you use a 600 dpi scanner, scanning at 1200 dpi will help you get rid of the moiré.

The second step in removing the patterns is to use effects like Gaussian Blur. Experiment with the effect radius to minimize the image quality loss (Gaussian Blur just blurs the image resulting in loss of sharpness). You can use not only integer values but also fractional. If the radius of 1 pixel seems to be insufficient but 2 pixels are too much, you can specify 1.5 pixels or thereabouts. Punching up the overall sharpness of the image, Gaussian Blur effect reduces fine jitter and speckles. However it can't heal the image from moirés completely.

Sometimes, scanning the original rotated to some angle (for example 15°) then rotating it back in an image editing software in conjunction with blur effect may further decrease the affect of moiré patterns.

Obviously the best way to remove the moiré patterns is to avoid scanning a printed originals. Always choose the photographs and slides when possible. Scan magazines and books only if no alternative exists.


[ DTP & GraphicsOberon Home Page ]
Copyright © 2000 by Alex Vakulenko. All rights reserved.
This page was last revised on 11/16/02.